Abstract: Reliability problems in fault detection for large-scale complicated systems are studied by using a probabilistic signed directed graph (SDG) model.Influences of sensor selection and location on reliability with respect to different sensor fault probabilities are analyzed.When system faults occur, the sensors can detect the departure of variances, but the system faults may be ignored or misdiagnosed because of the sensor faults, directly affecting the fault detection reliability. With description of the probabilistic SDG model, the reliability description of fault detection is given.Optimal sensor lo cation and the corresponding algorithm are studied under restricted resources.Furthermore, a probabilistic SDG model is established for a typical 65 t/h boiler, on which the above algorithm is executed to determine an optimization scheme. Simulation experiment shows that sensor location solved in this way is effective in improving fault detection reliability.

Key words: reliability, fault detection, sensor location, signed directed graph (SDG)